The discovery and mapping of cell types used to be slow and painstaking work but with the advent of single-cell transcriptomic techniques, such as single-cell RNA sequencing, this task can now be accomplished with relative ease. It is possible to discover and comprehensively catalog cell types from essentially any tissue while measuring, genome-wide, the molecular expression that gives rise to cellular function. This exciting ability is driving transformative, multi-agency efforts to catalog all cell types in humans and other model organisms, and these efforts promise to expand our understanding of tissue function both in health as well as in diseases such as cancer. However, the ambition of these atlas efforts still exceeds the capabilities of existing techniques: single- cell RNA sequencing provides limited insight into the spatial organization of tissues and can only characterize a small fraction of the tens to hundreds of millions of cells within even relatively small tissues. By contrast, image- based approaches to single-cell transcriptomics directly image a large fraction of the expressed transcriptome within cells in their native tissue context and, thus, provide spatial context to single-cell measurements; however, image-based approaches also do not have the throughput needed to characterize the substantial tissue volumes or cell numbers demanded by cell atlases efforts and more broad biological endeavors. As such, no current technique can meet the throughput demands of these ambitious cell atlas efforts, and the transformative promise they offer for expanding our understanding of disease remain largely out of reach. Here we will meet the demand for ultra-high-throughput single-cell techniques by substantially improving the throughput of an image-based approach that we helped pioneer?multiplexed error robust fluorescence in situ hybridization (MERFISH). This technique simultaneously identifies hundreds to thousands of different RNA molecules in single cells using single-molecule fluorescence imaging and can characterize tens of thousands of cells per day across ~mm3 tissue volumes. We will increase the throughput of this technique by two orders of magnitude, allowing measurements of thousands of genes within ~5 million cells per day across ~100 mm3 tissue volumes. We will achieve this increase in imaging bandwidth by combining 1) a novel camera bank system that uses 16 cameras rather than 1 per microscope with 2) an improved DNA-origami-based amplification system that will increase the brightness of single-molecules by ~100-fold to allow faster frame rates. In parallel, we will develop the improved microscope-control and analysis software and the data-transfer hardware required for the rapid transfer, long-term storage, and fast analysis of the tens of TB of data produced per day by this new imaging platform. Together the advances we propose will create a single-cell transcriptomics technique that is unrivaled in its ability to characterize massive numbers of cells across sizeable tissue volumes and which promises to enable a broad range of biological investigations, including cell atlas efforts, and, in the process, reveal the molecular and cellular basis of a diverse range of tissue functions.